# How to monitor indexation health with an AI agent

> Tier: deep. Full contract: quantified decision rules and evidence requirements.

Sitemaps, index coverage and freshness for the pages that earn money, with the drifts caught before they become traffic losses.

**Intent.** Keep the revenue pages indexed and freshly crawled, and catch coverage regressions while they are still cheap.

## When to use

- Weekly on large or fast-changing sites.
- After deploys touching routing, canonicals or sitemaps.
- When organic-decay-triage points at coverage.

## When not to use

- Small stable sites: quarterly is plenty, weekly is noise.
- As a crawler substitute: this reads Google's view via Search Console, not the site's own logs.

## Sources

Required: Search Console

## Preconditions

- Sitemaps submitted and the property ticked.
- The money-page list known, or derivable from clicks.

## Inputs

- `watchlist`: URLs that must stay indexed. Default: top pages by organic clicks.

## Procedure

1. Read sitemap health: submitted, read, errors, last processed. — tools: `gsc_list_sitemaps`, `gsc_get_sitemap_health`
2. Track sitemap deltas since last run: what appeared, what vanished. — tools: `gsc_track_sitemap_deltas`
3. Inspect the watchlist URLs in bulk: indexed, canonical, last crawl. — tools: `gsc_indexation_watchlist`, `gsc_bulk_inspect_urls`
4. Check crawl freshness against each page's change cadence. — tools: `gsc_monitor_indexation_freshness`
5. On very large sites, sample sections instead of exhausting quota. — tools: `gsc_plan_large_site_sampling`

## Decision rules

- **When** A money page reports indexed with a Google-selected canonical different from the declared one → **conclude** Canonical conflict: the page competes with itself; highest priority.
- **When** Sitemap URLs dropped sharply since the last delta → **conclude** Generation problem upstream; quote the counts and the sitemap file.
- **When** A watchlist page's last crawl exceeds twice its change cadence → **conclude** Staleness finding: Google serves an old version of a page that changed.
- **When** Coverage errors concentrate on one URL pattern → **conclude** One routing or template cause; name the pattern, not each URL.

## Evidence requirements

- Every finding quotes the inspection result verbatim: state, canonical, crawl date.
- Deltas carry both counts and the file they came from.

## Human approval boundary

- Fixes live in the site's code and infrastructure; the skill delivers the report.
- Requesting indexing at scale is rate-limited by Google and stays a human call.

## Failure modes

- Inspection quotas cap bulk checks; on big sites the read is a sample and must say so.
- Search Console lags real crawling by days; a deploy today judged tomorrow reads as broken.
- Intentional noindex pages inflate error counts; the watchlist must encode intent, not just traffic.

## Output

- **Watchlist status**: Each money page: indexed, canonical, last crawl.
- **Regressions**: New coverage problems since last run, by pattern.
- **Sitemap report**: Health, deltas, processing dates.

## Related

- Skill: [organic-decay-triage](https://www.getmcpads.com/skills/organic-decay-triage.md)
- Guide: [search-console-analysis](https://www.getmcpads.com/guides/search-console-analysis.md)